maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity
llvm-project/clang/test/CodeGen/builtins-ppc-quadword.c

250 lines
10 KiB
C
Raw Normal View History

This patch adds support for the vector quadword add/sub instructions introduced in POWER8. These are the Clang-related changes for http://reviews.llvm.org/D9081 vadduqm vaddeuqm vaddcuq vaddecuq vsubuqm vsubeuqm vsubcuq vsubecuq All builtins are added in altivec.h, and guarded with the POWER8_VECTOR and powerpc64 macros. http://reviews.llvm.org/D9903 llvm-svn: 238145
2015-05-25 23:52:45 +08:00
// REQUIRES: powerpc-registered-target
// RUN: %clang_cc1 -faltivec -target-feature +power8-vector \
// RUN: -triple powerpc64-unknown-unknown -emit-llvm %s -o - | FileCheck %s
// RUN: %clang_cc1 -faltivec -target-feature +power8-vector \
// RUN: -triple powerpc64le-unknown-unknown -emit-llvm %s -o - \
// RUN: | FileCheck %s -check-prefix=CHECK-LE
// RUN: not %clang_cc1 -faltivec -triple powerpc-unknown-unknown \
// RUN: -emit-llvm %s -o - 2>&1 | FileCheck %s -check-prefix=CHECK-PPC
The time when -faltivec (or, on clang only, -maltivec) will magically include altivec.h has come and gone. Rationale: This causes modules, rewrite-includes, etc to be sad and people should just include altivec.h in their source. llvm-svn: 264235
2016-03-24 09:26:08 +08:00
#include <altivec.h>
This patch adds support for the vector quadword add/sub instructions introduced in POWER8. These are the Clang-related changes for http://reviews.llvm.org/D9081 vadduqm vaddeuqm vaddcuq vaddecuq vsubuqm vsubeuqm vsubcuq vsubecuq All builtins are added in altivec.h, and guarded with the POWER8_VECTOR and powerpc64 macros. http://reviews.llvm.org/D9903 llvm-svn: 238145
2015-05-25 23:52:45 +08:00
// CHECK-PPC: error: __int128 is not supported on this target
vector signed __int128 vlll = { -1 };
// CHECK-PPC: error: __int128 is not supported on this target
vector unsigned __int128 vulll = { 1 };
[PowerPC] Implement plain VSX load/store builtins. Implement all the different 24 overloads for vec_xl and vec_xst. llvm-svn: 286455
2016-11-10 22:39:56 +08:00
signed long long param_sll;
// CHECK-PPC: error: __int128 is not supported on this target
signed __int128 param_lll;
// CHECK-PPC: error: __int128 is not supported on this target
unsigned __int128 param_ulll;
This patch adds support for the vector quadword add/sub instructions introduced in POWER8. These are the Clang-related changes for http://reviews.llvm.org/D9081 vadduqm vaddeuqm vaddcuq vaddecuq vsubuqm vsubeuqm vsubcuq vsubecuq All builtins are added in altivec.h, and guarded with the POWER8_VECTOR and powerpc64 macros. http://reviews.llvm.org/D9903 llvm-svn: 238145
2015-05-25 23:52:45 +08:00
// CHECK-PPC: error: __int128 is not supported on this target
vector signed __int128 res_vlll;
// CHECK-PPC: error: __int128 is not supported on this target
vector unsigned __int128 res_vulll;
// CHECK-LABEL: define void @test1
void test1() {
/* vec_add */
res_vlll = vec_add(vlll, vlll);
// CHECK: add <1 x i128>
// CHECK-LE: add <1 x i128>
// CHECK-PPC: error: call to 'vec_add' is ambiguous
res_vulll = vec_add(vulll, vulll);
// CHECK: add <1 x i128>
// CHECK-LE: add <1 x i128>
// CHECK-PPC: error: call to 'vec_add' is ambiguous
/* vec_vadduqm */
res_vlll = vec_vadduqm(vlll, vlll);
// CHECK: add <1 x i128>
// CHECK-LE: add <1 x i128>
// CHECK-PPC: error: assigning to '__vector __int128' (vector of 1 '__int128' value) from incompatible type 'int'
res_vulll = vec_vadduqm(vulll, vulll);
// CHECK: add <1 x i128>
// CHECK-LE: add <1 x i128>
// CHECK-PPC: error: assigning to '__vector unsigned __int128' (vector of 1 'unsigned __int128' value) from incompatible type 'int'
/* vec_vaddeuqm */
res_vlll = vec_vaddeuqm(vlll, vlll, vlll);
// CHECK: @llvm.ppc.altivec.vaddeuqm
// CHECK-LE: @llvm.ppc.altivec.vaddeuqm
// CHECK-PPC: error: assigning to '__vector __int128' (vector of 1 '__int128' value) from incompatible type 'int'
res_vulll = vec_vaddeuqm(vulll, vulll, vulll);
// CHECK: @llvm.ppc.altivec.vaddeuqm
// CHECK-LE: @llvm.ppc.altivec.vaddeuqm
// CHECK-PPC: error: assigning to '__vector unsigned __int128' (vector of 1 'unsigned __int128' value) from incompatible type 'int'
/* vec_addc */
res_vlll = vec_addc(vlll, vlll);
// CHECK: @llvm.ppc.altivec.vaddcuq
// CHECK-LE: @llvm.ppc.altivec.vaddcuq
// KCHECK-PPC: error: call to 'vec_addc' is ambiguous
res_vulll = vec_addc(vulll, vulll);
// CHECK: @llvm.ppc.altivec.vaddcuq
// CHECK-LE: @llvm.ppc.altivec.vaddcuq
// KCHECK-PPC: error: call to 'vec_addc' is ambiguous
/* vec_vaddcuq */
res_vlll = vec_vaddcuq(vlll, vlll);
// CHECK: @llvm.ppc.altivec.vaddcuq
// CHECK-LE: @llvm.ppc.altivec.vaddcuq
// CHECK-PPC: error: assigning to '__vector __int128' (vector of 1 '__int128' value) from incompatible type 'int'
res_vulll = vec_vaddcuq(vulll, vulll);
// CHECK: @llvm.ppc.altivec.vaddcuq
// CHECK-LE: @llvm.ppc.altivec.vaddcuq
// CHECK-PPC: error: assigning to '__vector unsigned __int128' (vector of 1 'unsigned __int128' value) from incompatible type 'int'
/* vec_vaddecuq */
res_vlll = vec_vaddecuq(vlll, vlll, vlll);
// CHECK: @llvm.ppc.altivec.vaddecuq
// CHECK-LE: @llvm.ppc.altivec.vaddecuq
// CHECK-PPC: error: assigning to '__vector __int128' (vector of 1 '__int128' value) from incompatible type 'int'
res_vulll = vec_vaddecuq(vulll, vulll, vulll);
// CHECK: @llvm.ppc.altivec.vaddecuq
// CHECK-LE: @llvm.ppc.altivec.vaddecuq
// CHECK-PPC: error: assigning to '__vector unsigned __int128' (vector of 1 'unsigned __int128' value) from incompatible type 'int'
/* vec_sub */
res_vlll = vec_sub(vlll, vlll);
// CHECK: sub <1 x i128>
// CHECK-LE: sub <1 x i128>
// CHECK-PPC: error: call to 'vec_sub' is ambiguous
res_vulll = vec_sub(vulll, vulll);
// CHECK: sub <1 x i128>
// CHECK-LE: sub <1 x i128>
// CHECK-PPC: error: call to 'vec_sub' is ambiguous
/* vec_vsubuqm */
res_vlll = vec_vsubuqm(vlll, vlll);
// CHECK: sub <1 x i128>
// CHECK-LE: sub <1 x i128>
// CHECK-PPC: error: assigning to '__vector __int128' (vector of 1 '__int128' value) from incompatible type 'int'
res_vulll = vec_vsubuqm(vulll, vulll);
// CHECK: sub <1 x i128>
// CHECK-LE: sub <1 x i128>
// CHECK-PPC: error: assigning to '__vector unsigned __int128' (vector of 1 'unsigned __int128' value) from incompatible type 'int'
/* vec_vsubeuqm */
res_vlll = vec_vsubeuqm(vlll, vlll, vlll);
// CHECK: @llvm.ppc.altivec.vsubeuqm
// CHECK-LE: @llvm.ppc.altivec.vsubeuqm
// CHECK-PPC: error: assigning to '__vector __int128' (vector of 1 '__int128' value) from incompatible type 'int'
[PPC] support for arithmetic builtins in the FE (commit again after fixing the buildbot failures) This adds various overloads of the following builtins to altivec.h: vec_neg vec_nabs vec_adde vec_addec vec_sube vec_subec vec_subc Note that for vec_sub builtins on 32 bit integers, the semantics is similar to what ISA describes for instructions like vsubecuq that work on quadwords: the first operand is added to the one's complement of the second operand. (As opposed to two's complement which I expected). llvm-svn: 287795
2016-11-24 02:36:29 +08:00
/* vec_sube */
res_vlll = vec_sube(vlll, vlll, vlll);
// CHECK: @llvm.ppc.altivec.vsubeuqm
// CHECK-LE: @llvm.ppc.altivec.vsubeuqm
// CHECK-PPC: error: call to 'vec_sube' is ambiguous
res_vulll = vec_sube(vulll, vulll, vulll);
// CHECK: @llvm.ppc.altivec.vsubeuqm
// CHECK-LE: @llvm.ppc.altivec.vsubeuqm
// CHECK-PPC: error: call to 'vec_sube' is ambiguous
res_vlll = vec_sube(vlll, vlll, vlll);
// CHECK: @llvm.ppc.altivec.vsubeuqm
// CHECK-LE: @llvm.ppc.altivec.vsubeuqm
// CHECK-PPC: error: call to 'vec_sube' is ambiguous
This patch adds support for the vector quadword add/sub instructions introduced in POWER8. These are the Clang-related changes for http://reviews.llvm.org/D9081 vadduqm vaddeuqm vaddcuq vaddecuq vsubuqm vsubeuqm vsubcuq vsubecuq All builtins are added in altivec.h, and guarded with the POWER8_VECTOR and powerpc64 macros. http://reviews.llvm.org/D9903 llvm-svn: 238145
2015-05-25 23:52:45 +08:00
res_vulll = vec_vsubeuqm(vulll, vulll, vulll);
// CHECK: @llvm.ppc.altivec.vsubeuqm
// CHECK-LE: @llvm.ppc.altivec.vsubeuqm
// CHECK-PPC: error: assigning to '__vector unsigned __int128' (vector of 1 'unsigned __int128' value) from incompatible type 'int'
[PPC] support for arithmetic builtins in the FE (commit again after fixing the buildbot failures) This adds various overloads of the following builtins to altivec.h: vec_neg vec_nabs vec_adde vec_addec vec_sube vec_subec vec_subc Note that for vec_sub builtins on 32 bit integers, the semantics is similar to what ISA describes for instructions like vsubecuq that work on quadwords: the first operand is added to the one's complement of the second operand. (As opposed to two's complement which I expected). llvm-svn: 287795
2016-11-24 02:36:29 +08:00
res_vulll = vec_sube(vulll, vulll, vulll);
// CHECK: @llvm.ppc.altivec.vsubeuqm
// CHECK-LE: @llvm.ppc.altivec.vsubeuqm
// CHECK-PPC: error: call to 'vec_sube' is ambiguous
This patch adds support for the vector quadword add/sub instructions introduced in POWER8. These are the Clang-related changes for http://reviews.llvm.org/D9081 vadduqm vaddeuqm vaddcuq vaddecuq vsubuqm vsubeuqm vsubcuq vsubecuq All builtins are added in altivec.h, and guarded with the POWER8_VECTOR and powerpc64 macros. http://reviews.llvm.org/D9903 llvm-svn: 238145
2015-05-25 23:52:45 +08:00
/* vec_subc */
res_vlll = vec_subc(vlll, vlll);
// CHECK: @llvm.ppc.altivec.vsubcuq
// CHECK-LE: @llvm.ppc.altivec.vsubcuq
// KCHECK-PPC: error: call to 'vec_subc' is ambiguous
res_vulll = vec_subc(vulll, vulll);
// CHECK: @llvm.ppc.altivec.vsubcuq
// CHECK-LE: @llvm.ppc.altivec.vsubcuq
// KCHECK-PPC: error: call to 'vec_subc' is ambiguous
/* vec_vsubcuq */
res_vlll = vec_vsubcuq(vlll, vlll);
// CHECK: @llvm.ppc.altivec.vsubcuq
// CHECK-LE: @llvm.ppc.altivec.vsubcuq
// CHECK-PPC: error: assigning to '__vector __int128' (vector of 1 '__int128' value) from incompatible type 'int'
res_vulll = vec_vsubcuq(vulll, vulll);
// CHECK: @llvm.ppc.altivec.vsubcuq
// CHECK-LE: @llvm.ppc.altivec.vsubcuq
// CHECK-PPC: error: assigning to '__vector unsigned __int128' (vector of 1 'unsigned __int128' value) from incompatible type 'int'
/* vec_vsubecuq */
res_vlll = vec_vsubecuq(vlll, vlll, vlll);
// CHECK: @llvm.ppc.altivec.vsubecuq
// CHECK-LE: @llvm.ppc.altivec.vsubecuq
[PPC] support for arithmetic builtins in the FE (commit again after fixing the buildbot failures) This adds various overloads of the following builtins to altivec.h: vec_neg vec_nabs vec_adde vec_addec vec_sube vec_subec vec_subc Note that for vec_sub builtins on 32 bit integers, the semantics is similar to what ISA describes for instructions like vsubecuq that work on quadwords: the first operand is added to the one's complement of the second operand. (As opposed to two's complement which I expected). llvm-svn: 287795
2016-11-24 02:36:29 +08:00
// CHECK-PPC: error: assigning to '__vector __int128' (vector of 1 '__int128' value) from incompatible type 'int'
[PPC] support for arithmetic builtins in the FE This adds various overloads of the following builtins to altivec.h: vec_neg vec_nabs vec_adde vec_addec vec_sube vec_subec vec_subc Note that for vec_sub builtins on 32 bit integers, the semantics is similar to what ISA describes for instructions like vsubecuq that work on quadwords: the first operand is added to the one's complement of the second operand. (As opposed to two's complement which I expected). llvm-svn: 287772
2016-11-24 00:32:05 +08:00
[PPC] Reverting r287772 Due to buildbot failure, I revert. Will recommit after investigation. llvm-svn: 287775
2016-11-24 00:56:03 +08:00
res_vulll = vec_vsubecuq(vulll, vulll, vulll);
[PPC] support for arithmetic builtins in the FE This adds various overloads of the following builtins to altivec.h: vec_neg vec_nabs vec_adde vec_addec vec_sube vec_subec vec_subc Note that for vec_sub builtins on 32 bit integers, the semantics is similar to what ISA describes for instructions like vsubecuq that work on quadwords: the first operand is added to the one's complement of the second operand. (As opposed to two's complement which I expected). llvm-svn: 287772
2016-11-24 00:32:05 +08:00
// CHECK: @llvm.ppc.altivec.vsubecuq
// CHECK-LE: @llvm.ppc.altivec.vsubecuq
[PPC] support for arithmetic builtins in the FE (commit again after fixing the buildbot failures) This adds various overloads of the following builtins to altivec.h: vec_neg vec_nabs vec_adde vec_addec vec_sube vec_subec vec_subc Note that for vec_sub builtins on 32 bit integers, the semantics is similar to what ISA describes for instructions like vsubecuq that work on quadwords: the first operand is added to the one's complement of the second operand. (As opposed to two's complement which I expected). llvm-svn: 287795
2016-11-24 02:36:29 +08:00
// CHECK-PPC: error: assigning to '__vector unsigned __int128' (vector of 1 'unsigned __int128' value) from incompatible type 'int'
res_vlll = vec_subec(vlll, vlll, vlll);
// CHECK: @llvm.ppc.altivec.vsubecuq
// CHECK-LE: @llvm.ppc.altivec.vsubecuq
// CHECK-PPC: error: assigning to '__vector __int128' (vector of 1 '__int128' value) from incompatible type 'int'
res_vulll = vec_subec(vulll, vulll, vulll);
// CHECK: @llvm.ppc.altivec.vsubecuq
// CHECK-LE: @llvm.ppc.altivec.vsubecuq
This patch adds support for the vector quadword add/sub instructions introduced in POWER8. These are the Clang-related changes for http://reviews.llvm.org/D9081 vadduqm vaddeuqm vaddcuq vaddecuq vsubuqm vsubeuqm vsubcuq vsubecuq All builtins are added in altivec.h, and guarded with the POWER8_VECTOR and powerpc64 macros. http://reviews.llvm.org/D9903 llvm-svn: 238145
2015-05-25 23:52:45 +08:00
// CHECK-PPC: error: assigning to '__vector unsigned __int128' (vector of 1 'unsigned __int128' value) from incompatible type 'int'
[PPC] add vector byte reverse functions to altivec.h This patch corresponds to review https://reviews.llvm.org/D25915. Committing on behalf of Sean Fertile. llvm-svn: 285268
2016-10-27 14:23:57 +08:00
res_vulll = vec_revb(vulll);
// CHECK: store <16 x i8> <i8 15, i8 14, i8 13, i8 12, i8 11, i8 10, i8 9, i8 8, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>, <16 x i8>* {{%.+}}, align 16
// CHECK: call <4 x i32> @llvm.ppc.altivec.vperm(<4 x i32> {{%.+}}, <4 x i32> {{%.+}}, <16 x i8> {{%.+}})
// CHECK-LE: store <16 x i8> <i8 15, i8 14, i8 13, i8 12, i8 11, i8 10, i8 9, i8 8, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>, <16 x i8>* {{%.+}}, align 16
// CHECK-LE: store <16 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, <16 x i8>* {{%.+}}, align 16
// CHECK-LE: xor <16 x i8>
// CHECK-LE: call <4 x i32> @llvm.ppc.altivec.vperm(<4 x i32> {{%.+}}, <4 x i32> {{%.+}}, <16 x i8> {{%.+}})
// CHECK_PPC: error: call to 'vec_revb' is ambiguous
[PowerPC] Implement plain VSX load/store builtins. Implement all the different 24 overloads for vec_xl and vec_xst. llvm-svn: 286455
2016-11-10 22:39:56 +08:00
/* vec_xl */
res_vlll = vec_xl(param_sll, &param_lll);
// CHECK: load <1 x i128>, <1 x i128>* %{{[0-9]+}}, align 16
// CHECK-LE: load <1 x i128>, <1 x i128>* %{{[0-9]+}}, align 16
// CHECK-PPC: error: call to 'vec_xl' is ambiguous
res_vulll = vec_xl(param_sll, &param_ulll);
// CHECK: load <1 x i128>, <1 x i128>* %{{[0-9]+}}, align 16
// CHECK-LE: load <1 x i128>, <1 x i128>* %{{[0-9]+}}, align 16
// CHECK-PPC: error: call to 'vec_xl' is ambiguous
/* vec_xst */
vec_xst(vlll, param_sll, &param_lll);
// CHECK: store <1 x i128> %{{[0-9]+}}, <1 x i128>* %{{[0-9]+}}, align 16
// CHECK-LE: store <1 x i128> %{{[0-9]+}}, <1 x i128>* %{{[0-9]+}}, align 16
// CHECK-PPC: error: call to 'vec_xst' is ambiguous
vec_xst(vulll, param_sll, &param_ulll);
// CHECK: store <1 x i128> %{{[0-9]+}}, <1 x i128>* %{{[0-9]+}}, align 16
// CHECK-LE: store <1 x i128> %{{[0-9]+}}, <1 x i128>* %{{[0-9]+}}, align 16
// CHECK-PPC: error: call to 'vec_xst' is ambiguous
[PowerPC] Implement BE VSX load/store builtins - clang portion. This patch implements all the overloads for vec_xl_be and vec_xst_be. On BE, they behaves exactly the same with vec_xl and vec_xst, therefore they are simply implemented by defining a matching macro. On LE, they are implemented by defining new builtins and intrinsics. For int/float/long long/double, it is just a load (lxvw4x/lxvd2x) or store(stxvw4x/stxvd2x). For char/char/short, we also need some extra shuffling before or after call the builtins to get the desired BE order. For int128, simply call vec_xl or vec_xst. llvm-svn: 286971
2016-11-15 22:30:56 +08:00
/* vec_xl_be */
res_vlll = vec_xl_be(param_sll, &param_lll);
// CHECK: load <1 x i128>, <1 x i128>* %{{[0-9]+}}, align 16
// CHECK-LE: load <1 x i128>, <1 x i128>* %{{[0-9]+}}, align 16
// CHECK-PPC: error: call to 'vec_xl' is ambiguous
res_vulll = vec_xl_be(param_sll, &param_ulll);
// CHECK: load <1 x i128>, <1 x i128>* %{{[0-9]+}}, align 16
// CHECK-LE: load <1 x i128>, <1 x i128>* %{{[0-9]+}}, align 16
// CHECK-PPC: error: call to 'vec_xl' is ambiguous
/* vec_xst_be */
vec_xst_be(vlll, param_sll, &param_lll);
// CHECK: store <1 x i128> %{{[0-9]+}}, <1 x i128>* %{{[0-9]+}}, align 16
// CHECK-LE: store <1 x i128> %{{[0-9]+}}, <1 x i128>* %{{[0-9]+}}, align 16
// CHECK-PPC: error: call to 'vec_xst' is ambiguous
vec_xst_be(vulll, param_sll, &param_ulll);
// CHECK: store <1 x i128> %{{[0-9]+}}, <1 x i128>* %{{[0-9]+}}, align 16
// CHECK-LE: store <1 x i128> %{{[0-9]+}}, <1 x i128>* %{{[0-9]+}}, align 16
// CHECK-PPC: error: call to 'vec_xst' is ambiguous
This patch adds support for the vector quadword add/sub instructions introduced in POWER8. These are the Clang-related changes for http://reviews.llvm.org/D9081 vadduqm vaddeuqm vaddcuq vaddecuq vsubuqm vsubeuqm vsubcuq vsubecuq All builtins are added in altivec.h, and guarded with the POWER8_VECTOR and powerpc64 macros. http://reviews.llvm.org/D9903 llvm-svn: 238145
2015-05-25 23:52:45 +08:00
}